feat(agent): 24 fournisseurs LLM + refonte du panneau Agent & IA

Enregistre OpenAI, Anthropic, Mistral, Cohere, Google Gemini, Groq, DeepSeek, OpenRouter, NVIDIA NIM, Together, Perplexity, xAI, DashScope, MiniMax, Morph, Fireworks, Cerebras, SambaNova, Chutes, Xiaomi, LTX, SEA-LION, SenseNova et MemTensor (URLs de base + presets de modeles + libelles). Settings: panneau Agent & IA repense en maitre/detail (liste rechercheable + volet de configuration) pour tenir avec des dizaines de fournisseurs, en conservant tous les boutons/fonctions.
2026-09-14 22:25:57 -04:00
parent 0d475c3d2d
commit fb14c7e709
4 changed files with 225 additions and 61 deletions
@@ -5,8 +5,11 @@ directly — it emits *tool intentions* (function calls) that AgentEngine turns
 into guarded internal actions.

 Supported providers (OpenAI-compatible chat-completions JSON response):
-  openai, anthropic*, google*, deepseek, qwencloud, nvidia, openrouter, ollama.
-  (* routed through an OpenAI-compatible gateway / any configured api_base)
+  openai, anthropic, mistral, cohere, google, groq, deepseek, openrouter,
+  nvidia, together, perplexity, xai, qwencloud, minimax, morph, fireworks,
+  cerebras, sambanova, chutes, xiaomi, ltx, sealion, sensenova, memtensor,
+  ollama (local, no key). Anthropic/Google/Cohere expose an OpenAI-compatible
+  surface at their configured base URL.

 When no API key is configured (or provider == "offline") the client falls back
 to a deterministic, dependency-free *mock planner*. This keeps the whole agent
@@ -28,30 +31,106 @@ from app.config import settings
 logger = logging.getLogger(__name__)

 # Provider → default model + base URL when llm_model/api_base are empty.
+# All entries speak the OpenAI-compatible chat-completions protocol (Anthropic,
+# Google and Cohere expose an OpenAI-compatible surface at their given base).
 PROVIDERS = {
    "openai":     ("https://api.openai.com/v1",      "gpt-4o"),
    "anthropic":  ("https://api.anthropic.com/v1",   "claude-opus-4-8"),
+    "mistral":    ("https://api.mistral.ai/v1",      "mistral-large-latest"),
+    "cohere":     ("https://api.cohere.com/v2",      "command-r-plus"),
    "google":     ("https://generativelanguage.googleapis.com/v1beta", "gemini-2.0-pro"),
+    "groq":       ("https://api.groq.com/openai/v1", "llama-3.3-70b-versatile"),
    "deepseek":   ("https://api.deepseek.com/v1",    "deepseek-chat"),
-    "qwencloud":  ("https://dashscope.aliyuncs.com/compatible-mode/v1", "qwen-max"),
-    "nvidia":     ("https://integrate.api.nvidia.com/v1", "nvidia/nemotron-3-super-120b-a12b"),
    "openrouter": ("https://openrouter.ai/api/v1",   "meta-llama/llama-3.3-70b-instruct"),
+    "nvidia":     ("https://integrate.api.nvidia.com/v1", "nvidia/nemotron-3-super-120b-a12b"),
+    "together":   ("https://api.together.xyz/v1",    "meta-llama/Llama-3.3-70B-Instruct-Turbo"),
+    "perplexity": ("https://api.perplexity.ai/v1",   "sonar-pro"),
+    "xai":        ("https://api.x.ai/v1",            "grok-2-latest"),
+    "qwencloud":  ("https://dashscope-intl.aliyuncs.com/compatible-mode/v1", "qwen-max"),
+    "minimax":    ("https://api.minimax.chat/v1",    "MiniMax-Text-01"),
+    "morph":      ("https://api.morphllm.com/v1",    "morph-v3-large"),
+    "fireworks":  ("https://api.fireworks.ai/inference/v1",
+                   "accounts/fireworks/models/llama-v3p3-70b-instruct"),
+    "cerebras":   ("https://api.cerebras.ai/v1",     "llama-3.3-70b"),
+    "sambanova":  ("https://api.sambanova.ai/v1",    "Meta-Llama-3.3-70B-Instruct"),
+    "chutes":     ("https://api.chutes.ai/v1",       "deepseek-ai/DeepSeek-V3"),
+    "xiaomi":     ("https://api.xiaomimimo.com/v1",  "mimo-7b-rl"),
+    "ltx":        ("https://api.ltx.io/v1",          "ltx-video"),
+    "sealion":    ("https://api.sea-lion.ai/v1",     "aisingapore/Llama-SEA-LION-v3-70B-IT"),
+    "sensenova":  ("https://token.sensenova.cn/v1",  "SenseChat-5"),
+    "memtensor":  ("https://memos.memtensor.cn/api/openmem/v1", "memos-chat"),
    "ollama":     ("http://localhost:11434/v1",      "llama3.1"),
    "offline":    (None, None),
 }

+# Friendly display names for the Settings / Agent UIs.
+PROVIDER_LABELS: dict[str, str] = {
+    "openai":     "OpenAI",
+    "anthropic":  "Anthropic",
+    "mistral":    "Mistral",
+    "cohere":     "Cohere",
+    "google":     "Google Gemini",
+    "groq":       "Groq",
+    "deepseek":   "DeepSeek",
+    "openrouter": "OpenRouter",
+    "nvidia":     "NVIDIA NIM",
+    "together":   "Together AI",
+    "perplexity": "Perplexity",
+    "xai":        "xAI (Grok)",
+    "qwencloud":  "DashScope (Alibaba)",
+    "minimax":    "MiniMax",
+    "morph":      "Morph",
+    "fireworks":  "Fireworks AI",
+    "cerebras":   "Cerebras",
+    "sambanova":  "SambaNova",
+    "chutes":     "Chutes AI",
+    "xiaomi":     "Xiaomi (MiMo)",
+    "ltx":        "LTX Studio",
+    "sealion":    "SEA-LION",
+    "sensenova":  "SenseNova",
+    "memtensor":  "MemTensor (MemOS)",
+    "ollama":     "Ollama (local)",
+    "offline":    "Hors-ligne (mock)",
+}
+
 # Curated model presets surfaced by /api/agent/providers for the UI selectors.
 PROVIDER_MODELS: dict[str, list[str]] = {
    "openai":     ["gpt-4o", "gpt-4o-mini", "gpt-4.1", "gpt-4.1-mini", "o3-mini", "gpt-4-turbo"],
    "anthropic":  ["claude-opus-4-8", "claude-sonnet-4-5", "claude-3-5-sonnet", "claude-haiku-4-5"],
+    "mistral":    ["mistral-large-latest", "mistral-medium-latest", "mistral-small-latest",
+                   "codestral-latest", "open-mistral-nemo", "pixtral-large-latest"],
+    "cohere":     ["command-r-plus", "command-r", "command-light", "command-a"],
    "google":     ["gemini-2.0-pro", "gemini-2.0-flash", "gemini-1.5-pro", "gemini-1.5-flash"],
+    "groq":       ["llama-3.3-70b-versatile", "llama-3.1-8b-instant",
+                   "mixtral-8x7b-32768", "gemma2-9b-it"],
    "deepseek":   ["deepseek-chat", "deepseek-reasoner"],
-    "qwencloud":  ["qwen-max", "qwen-plus", "qwen-turbo", "qwen-long"],
+    "openrouter": ["meta-llama/llama-3.3-70b-instruct", "anthropic/claude-3.5-sonnet",
+                   "openai/gpt-4o", "mistralai/mistral-large"],
    "nvidia":     ["nvidia/nemotron-3-super-120b-a12b", "nvidia/nemotron-3-nano-30b-a3b",
               "meta/llama-3.1-70b-instruct", "nvidia/llama-3.3-nemotron-super-49b-v1.5",
               "deepseek-ai/deepseek-v4-pro", "z-ai/glm-5.2"],
-    "openrouter": ["meta-llama/llama-3.3-70b-instruct", "anthropic/claude-3.5-sonnet",
-                   "openai/gpt-4o", "mistralai/mistral-large"],
+    "together":   ["meta-llama/Llama-3.3-70B-Instruct-Turbo",
+                   "meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo",
+                   "Qwen/Qwen2.5-72B-Instruct-Turbo", "mistralai/Mixtral-8x7B-Instruct-v0.1"],
+    "perplexity": ["sonar-pro", "sonar", "sonar-reasoning", "sonar-deep-research"],
+    "xai":        ["grok-2-latest", "grok-2", "grok-beta", "grok-3"],
+    "qwencloud":  ["qwen-max", "qwen-plus", "qwen-turbo", "qwen-long"],
+    "minimax":    ["MiniMax-Text-01", "abab6.5s-chat", "abab6.5-chat"],
+    "morph":      ["morph-v3-large", "morph-v3-fast"],
+    "fireworks":  ["accounts/fireworks/models/llama-v3p3-70b-instruct",
+                   "accounts/fireworks/models/mixtral-8x22b-instruct",
+                   "accounts/fireworks/models/qwen2p5-72b-instruct"],
+    "cerebras":   ["llama-3.3-70b", "llama3.1-8b", "llama-3.1-70b"],
+    "sambanova":  ["Meta-Llama-3.3-70B-Instruct", "Meta-Llama-3.1-405B-Instruct",
+                   "Qwen2.5-72B-Instruct"],
+    "chutes":     ["deepseek-ai/DeepSeek-V3", "deepseek-ai/DeepSeek-R1",
+                   "Qwen/Qwen2.5-72B-Instruct"],
+    "xiaomi":     ["mimo-7b-rl", "mimo-7b"],
+    "ltx":        ["ltx-video", "ltx-video-2"],
+    "sealion":    ["aisingapore/Llama-SEA-LION-v3-70B-IT",
+                   "aisingapore/Gemma-SEA-LION-v3-9B-IT"],
+    "sensenova":  ["SenseChat-5", "SenseChat-5-Cantonese", "SenseChat-Turbo"],
+    "memtensor":  ["memos-chat"],
    "ollama":     ["llama3.1", "llama3", "mistral", "qwen2.5", "gemma2", "mixtral"],
    "offline":    [],
 }
@@ -14,7 +14,7 @@ import json
 import time

 from app.config import settings
-from app.services.llm_client import PROVIDER_MODELS, PROVIDERS
+from app.services.llm_client import PROVIDER_LABELS, PROVIDER_MODELS, PROVIDERS

 # Providers whose /v1/models lists far more entries than /v1/chat/completions
 # actually serves. The fetched list is validated (name filter + live probe)
@@ -165,7 +165,7 @@ def provider_info() -> list[dict]:
            models.insert(0, default_model)
        out.append({
            "id": name,
-            "name": name.capitalize(),
+            "name": PROVIDER_LABELS.get(name) or name.replace("_", " ").title(),
            "default_model": default_model,
            "models": models,
            "base_url": (base or ""),